Page 1
Thermal Process Calculations Through Ball’s Original FormulaMethod: A Critical Presentation of the Method and Simplificationof its Use Through Regression Equations
Nikolaos G. Stoforos
Received: 10 August 2009 / Accepted: 8 February 2010 / Published online: 3 March 2010
� Springer Science+Business Media, LLC 2010
Abstract Ball’s formula method is a classical method
used in the thermal processing industry, the basis and the
precursor of a number of most recent methodologies. It has
received a lot of attention, being reviewed, criticized,
compared and evaluated by several investigators. The use
of Ball’s method relies on appropriate diagrams which
sometimes are difficult to use, find or reproduce. After
presenting the principles involved in the development of
Ball’s formula method, reviewing the literature related to
that particular method, and discussing its limitations, some
working, regression equations developed in order to facil-
itate the application and the use of the method are
presented.
Keywords Ball’s formula method � Thermal processing �Canned foods � Process calculations
Latin Letters
a1–a7 Regression coefficients appearing in Eq. 27,
dimensionless
b1–b7 Regression coefficients appearing in Eq. 28,
associated with temperature differences, g and z,
in �F
B Steam-off time (measured from corrected zero),
min
C Concentration of a heat-labile substance, number
of microorganisms/mL, spores per container,
g/mL, or any other appropriate unit
CUT Duration of retort come-up time, min
DT
(Noted also as D) decimal reduction time or death
rate constant; time at a constant temperature
required to reduce by 90% the initial spore load
(or, in general, time required for 90% reduction of
a heat-labile substance), min
E1 Exponential integral defined by Eq. 15
Ei Exponential integral defined by Eq. 16
FzT (Or simply F) time at a constant temperature, T,
required to destroy a given percentage of
microorganisms whose thermal resistance is
characterized by z, or, the equivalent processing
time of a hypothetical thermal process at a
constant temperature that produces the same
effect (in terms of spore destruction) as the
actual thermal process, min
Fi Factor defined by Eq. 14, which when multiplied
FTref by gives the F value at the retort temperature,
dimensionless
f Time required for the difference between the
medium and the product temperature to change by
a factor of 10 min
g Difference between retort and product temperature
(at the critical point) at steam-off time, �F
g0 A small g value lower than or close to 0.1�F, after
which product temperature is considered constant, �F
I Difference between retort and initial product
temperature, �F
j A correction factor defined by Eqs. 5 and 8 for the
heating and cooling curve, respectively, based on
the intercept, with the temperature axis at time
zero, of the straight line that describes the late,
straight, portion of the experimental heating or
cooling curve plotted in a semi-logarithmic
temperature difference scale as shown in Figs. 1
and 2, respectively, dimensionless
N. G. Stoforos (&)
Department of Food Science and Technology, Agricultural
University of Athens, Iera Odos 75, 11855 Athens, Greece
e-mail: [email protected]
123
Food Eng Rev (2010) 2:1–16
DOI 10.1007/s12393-010-9014-4
Page 2
m Difference between product temperature at steam-
off and retort temperature during cooling, defined
by Eq. 12, �F
T (Product) temperature, �F
TA Extrapolated pseudo-initial product temperature at
the beginning of heating defined as the intercept,
with the temperature axis at time zero, of the
straight line that describes the late, straight,
portion of the experimental heating curve plotted
as shown in Fig. 1, �F
TB Extrapolated pseudo-initial product temperature at
the beginning of cooling defined as the intercept,
with the temperature axis at zero cooling time, of
the straight line that describes the late, straight,
portion of the experimental cooling curve plotted
as shown in Fig. 2, �F
Th Product temperature at the beginning of the
cooling cycle, �F
t Time, min
U The F value at TRT, defined by Eq. 13, min
u Dummy variable
z Temperature difference required to achieve a
decimal change of the DT value, �F
zc A correction temperature difference factor
appearing in Eq. 27, �F
Latin Letters
q The fraction of the total lethal value of a process (that is,
the F value of the entire process) which is achieved during
the heating cycle only of the thermal process, assuming
that the slope of the heating curve is constant and equal to
the slope of the cooling curve, dimensionless
Subscripts
1, 2 Refers to a particular condition
a Initial condition
b Final condition
bh Condition at the time where the break, the
change in the slope of the heating curve occurs
CW (Water) cooling medium
c Cooling phase
end End of cooling cycle
g Condition at steam-off time
h Heating phase
IT Initial condition (for product temperature only)
process Referring to process values
RT (Retort) heating medium
ref Reference value
required Referring to required values
Introduction
Thermal process calculations refer to the design and/or the
evaluation of a thermal process in terms of its ability to
extend the shelf life of a product by destroying undesirable
agents present in the food while preserving the quality
characteristics of the product. The General Method [9] was
the first method developed for thermal process calculations.
It was the starting point for a scientific approach in the
production of canned products, and it set the fundamental
principles for subsequent developments.
The General Method is an exact method as far as the
evaluation of a thermal process is concerned, but it lacks
the predictive power needed for design purposes. The
method developed by C. Olin Ball and at first presented in
1923 not only made the tedious calculations required by the
General Method easier, but it successfully engineered the
thermal process design task. It was the first of the so called
Formula Methods, a classical method still used in the
thermal processing industry, recognized as ‘‘a major
milestone in the history of food technology’’ by Merson
et al. [31] in their evaluation of the method. It served as the
model for a number of methods, some of them directly
influenced being those by Ball and Olson, Herndon et al.,
Griffin et al., Hayakawa, Larkin, and Larkin and Berry
[6, 23, 18, 19, 20, 28 and 30]. Limited comparisons of
Heating Curve (T RT =250°F)
210
220
230
240241242243244245
246
247
248
249
200190180170160150
50
-50
-150
-250-350-450-550-650-750
TIT I=TRT-TIT
TA jhI=TRT-TA
0 20 40 60 80 100 120
Time (min)
Prod
uct T
empe
ratu
re, T
(°F
)
1
10
100
1000
Ret
ort -
Pro
duct
Tem
pera
ture
, TR
T -
T (
°F)
Exponential Fitting (to the "linear"portion of the experimental data)
Experimental Data
f h =48 min
2170
340 ==−−
=ITRT
ARTh TT
TTj
Fig. 1 Typical straight-line heating curve used for fh and jh parameter
estimation
2 Food Eng Rev (2010) 2:1–16
123
Page 3
various formula methods have been reported in the litera-
ture [17, 42, 47].
In this presentation, we will be concentrated to matters
directly associated with Ball’s original formula method. The
term ‘‘original’’ formula method is used here in reference to
the method presented initially by Ball in 1923 and thereafter in
subsequent publications [4, 6], and to distinguish from a
‘‘new’’ method included in the 1957 publication [6]. Reviews
on mathematical procedures developed for thermal process
calculations are available [10, 21, 50]. A variety of comput-
erized thermal process calculation methodologies based on
numerical methods [27, 33, 55], and the General Method [41]
have been also appeared in the literature with some of them
being for commercial use [24]. They are flexible enough to
adjust to different process conditions as well as to handle
process deviations. A review of such methods, stretching their
advantages and limitations, will be always useful.
The objective of this work was to give a critical pre-
sentation of Ball’s original formula method for thermal
process calculations and to discus the assumptions involved
in terms of product safety and product quality. In order to
facilitate the use of the method, a set of regression equa-
tions was developed and is included in this article.
Theoretical Considerations
The concept of the F value of a thermal process is the
foundation of thermal process calculations. The F value of
a process is defined as the equivalent time of a hypothetical
thermal process at constant temperature, Tref, that produces
the same result, as far as the destruction of microorganisms
or other spoilage agents is concerned, with the actual
thermal process, during which product temperature can be,
and usually is, nonconstant. In the remaining of the text,
when talking about thermal destruction we will be referring
to microbial destruction, although the analysis will apply to
any heat-labile substance. From the aforementioned defi-
nition of the F value, and accepting first order kinetics for
microbial thermal destruction [12], we can further define
the F value as the time, at a constant temperature, Tref,
required to destroy a given percentage of microorganisms.
Following the classical thermobacteriological approach for
first order destruction kinetics [6, 32, 51], the F value can
be mathematically expressed as:
FzTref¼ DTref
ðlogðCaÞ � logðCbÞÞ ¼Ztb
ta
10TðtÞ�Tref
z dt ð1Þ
Superscript z on the F value indicates that the F value of
a process is associated to a particular type of
microorganism or heat-labile substance characterized by
a given z value. More details and some discussion
concerning the above equation can be found elsewhere
[49]. In brief, Eq. 1 suggests evaluation of the F value at a
particular point in the product either through microbial
concentration data at that particular point at the beginning
and at the end of the process, or through the time
temperature data observed at the point of interest
throughout the entire process. Although the two modes of
calculation produce equivalent results, provided the
parameters involved are known with adequate accuracy,
it is customary to use microbial reduction data to establish
a required F value (Frequired) and temperature evolution
data to calculate the F value of a given process (Fprocess).
There are two type of problems associated with thermal
process calculations. The first one refers to the calculation
of the Fprocess value of a given process; that is, a process for
which we know the heating time and the time–temperature
history of the product throughout the total processing time.
The second problem refers to the calculation of the
required heating time, at given processing conditions, in
order to achieve a target Frequired value; an F value which is
Cooling Curve (T CW =70°F)
110
100
90
8079787776
75
74
73
72
71
120
130140150160170
270
370
470
570
670770870970
1070
Th Th-TCW
TB TB-TCW
0 20 40 60 80 100 120
Time (min)
Prod
uct T
empe
ratu
re, T
(°F
)
1
10
100
1000
Prod
uct -
Coo
ling
Wat
er T
empe
ratu
re, T
- T
CW
(°F
)
Exponential Fitting (to the "linear"portion of the experimental data)
Experimental Data
f c =48 min
41.17.172
5.243 ==−−
=CWh
CWBc TT
TTj
Fig. 2 Typical cooling curve used for fc and jc parameter estimation
Food Eng Rev (2010) 2:1–16 3
123
Page 4
based on predescribed microbial reduction requirements. If
we split the integral of Eq. 1 into two parts, the first one
referring to the heating cycle and the second one to the
cooling cycle (that inevitably follows the heating cycle) of
the thermal process, that is,
Ztb
ta
10TðtÞ�Tref
z dt ¼Zth¼B
th¼0
10TðthÞ�Tref
z dthþZtc¼tend
tc¼0
10TðtcÞ�Tref
z dtc ð2Þ
then, both types of problems can be attacked through the
solution of the same equation:
Rth¼B
th¼0
10TðthÞ�Tref
z dthþRtc¼tend
tc¼0
10TðtcÞ�Tref
z dtc
Fz
Tref
¼ 1 ð3Þ
If the heating time, B, of a given process is known, the
solution for Fz
Trefthrough Eq. 3 gives the Fprocess value. This
is the process evaluation step and, given the time–
temperature data of the product during the entire process,
it is rather straightforward. Alternatively, if Fz
Trefrepresents
the Frequired value, one can solve Eq. 3 for the heating time,
B, the upper limit of the first integral in the above equation.
This is the design step of a thermal process, a much more
difficult mathematical task, which also requires knowledge
of product time–temperature evolution data. As a matter of
fact, this step is the problematic step in the General Method
for thermal process calculations presented by Bigelow and
his colleagues in 1920 and which probably led Ball into the
development of his Formula Method. We must notice here
that an a priori knowledge of the F value that will be
accumulated after the end of the heating (i.e., during the
cooling cycle of the process) as it is represented here by the
second integral of Eq. 3 is a prerequisite for the calculation
of the heating time and thus for the process design step.
Ball’s Formula Method
As far as food safety, in terms of microbial destruction, is
concerned, Ball was a supporter of the single point F value
calculation theory. He reasoned that ‘‘bacterial spores are
most likely to survive in the unit volume which receives the
smallest amount of lethal heat’’ [5]. Thus, for conduction
heated foods for example, the geometric center of the
container or a region close to it [14, 54] represents the
critical point of the product; the point that is least affected
by the process. If the product at this point is ‘‘safe’’, then
the whole product is safe. Thermal process calculations, as
for example temperature data used with Eq. 3 for F value
calculations should be, therefore, based on the critical point
of the product.
The approach that Ball [3] used in developing his ori-
ginal method set the basis of a group of calculation
methods named Formula Methods [50]. He first developed
a set of equations to describe the temperature of the
product, at the critical point, as a function of processing
time and then, by substituting these equations into Eq. 3
and solving the resulted equation, he obtained a relation-
ship between heating time and F value. Actually, as it will
be presented later on, among a choice of independent and
depended variables, Ball expressed the results of integra-
tion of Eq. 3 through the dimensionless fh/U parameter and
the difference between retort and product temperature, at
the critical point, at the end of heating (g).
Based on observations of the temperature evolution at
the critical point of canned foods, Ball concluded that any
heating or cooling curve, when appropriately plotted, after
an initial lag was asymptote to one (straight-line curve) or
more (broken-line curve) straight lines. Such heating and
cooling curves, as traditionally plotted in thermal process
literature (for example, in ‘‘inverse’’ semi-logarithmic
retort minus product temperature difference scale, for the
heating curve) are shown in Figs. 1 and 2, respectively
(temperatures will be given in degrees Fahrenheit and time
in minutes, following the terminology in Ball’s publica-
tions and the units associated with the use of Ball’s
method). After the initial lag, each curve could be, there-
fore, approximated by its asymptote. So, for a straight-line
heating curve, he suggested the following equation:
TðthÞ ¼ TRT � jhðTRT � TITÞ10�th=fh ð4Þ
Equation 4 assumes constant retort temperature
throughout the heating cycle, and as it is evidence from
this equation, the entire heating curve was described
through two empirical parameters, the fh and jh values. The
fh value is related to the slope of the heating curve and it is
defined as the time required for the difference between
retort and product temperature to change by a factor of 10,
that is, in relation to Fig. 1, as the time needed for the
straight-line heating curve to traverse through a logarithmic
cycle. The jh value is a dimensionless lag factor [48]
defined as:
jh ¼TRT � TA
TRT � TITð5Þ
for TA being an extrapolated pseudo-initial product tem-
perature at the beginning of heating (Fig. 1). Equation 4
written as temperature difference ratios, that is, (TRT–T(th))
over (TRT–TIT) suggests the independency of the fh and jhvalues on the temperature units used.
Determination of the fh and jh parameters might intro-
duce some error in subsequent calculations, not necessarily
for their graphical determination as it was initially done
(computer software allows nowadays accurate data fitting
4 Food Eng Rev (2010) 2:1–16
123
Page 5
and parameter estimation) but because their values are
affected by the length of heating time during which heat
penetration data available [34]. Nevertheless, these two
empirical fh and jh parameters can be related, through
theoretical expressions, to product properties and process
parameters, so that they can be transferred from a given
experiment to a number of different product and process
conditions [6, 34].
Ball used Eq. 4 for the entire heating curve, reasoning
that the thermal destruction taking place at the beginning of
a thermal process, where Eq. 4 might not successfully
describe product temperature data, is negligible since,
under common commercial practices, the temperature of
the product has not yet reached the lethal temperature
range. Contrary to this, the cooling curve was described by
two equations: one for the initial curved portion of the data
and another for the straight-line part of the curve. At the
beginning of cooling, product temperature is high and this
must be accurately taken into account in order to arrive to
correct process calculations. So, Ball used a hyperbola to
describe the initial part of the cooling curve for
0 B tc B fclog(jc/0.657), as given here by Eq. 6:
TðtcÞ¼Tgþ0:3ðTg
�TCWÞ 1�
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ 1
0:5275logðjc=0:657Þ
� �2tc
fc
� �2s2
435
ð6Þ
where Tg is the temperature of the product at the critical
point at the end of heating, and TCW is the retort temper-
ature during the cooling cycle.
Based on experimental observations, Ball [3] made a
series of assumptions about the parameters defining the
hyperbola. He fixed the shape of the hyperbola through
various constants and parameters, which were based on the
difference between product temperature at the end of heating
(Tg) and the retort temperature during cooling (TCW).
For the straight-line portion of the cooling curve, that is, for
tc C fclog(jc/0.657), Ball used another exponential equation:
TðtcÞ ¼ TCW þ jcðTg � TCWÞ10�tc=fc ð7Þ
As for the heating cycle, the fc and jc parameters are
determined from the cooling curve (Fig. 2). In particular, jcis defined as:
jc ¼TB � TCW
Th � TCWð8Þ
In a next step, based on experimental observations and
conservative assumptions, Ball fixed the jc value at 1.41
and he further assumed fc = fh, being thus liberated from
the need of collecting cooling data. Under the new
assumptions, Eq. 6 is reduced to:
TðtcÞ ¼ Tg þ 0:3ðTg � TCWÞ 1�
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ tc
0:175fh
� �2s2
435
ð9Þ
Three main restrictions being so far imposed in Ball’s
formula method due to the assumptions made about
product temperature during the cooling cycle (Eq. 7 and
9) are as follows: First, the equation used to describe the
initial part of the cooling curve (Eq. 9) does not allow for
the product temperature to increase at the beginning of
cooling. This is not valid for the geometric center (or the
critical point) of conduction heated canned foods. Second,
a jc value of 1.41 can be rarely (and accidentally) true and
might be a severe limitation. A jc value of 1 is expected for
perfect mixing forced convection heated foods, while a
value of 2.04 is the theoretical value for conduction heated
canned foods at uniform initial temperature at the
beginning of cooling and processing conditions that result
in infinite heat transfer coefficient between product and
cooling medium. Third, the assumption about fc = fh is an
assumption that facilitates calculations. The case where the
slope of the heating curve is different from the slope of the
cooling curve (that is, the case when fc = fh) has been
discussed and accounted for by Ball [3]. The
aforementioned limitations will be further addressed in
this article.
Having described with the appropriate equations product
temperature evolution at the critical point for the entire
thermal process, Ball proceeded with the substitution of
these equations (Eqs. 4, 7 and 9) into Eq. 3 and, after
considerable mathematical operations, he ended up with
the following equation:
fh
U
1
lnð10Þ elnð10Þg=z
� elnð10Þg=z E1
lnð10Þ gz
� �� E1
lnð10Þ 80
z
� �� ��
þ 0:76381e�0:789m=z þ 0:5833z
me0:692m=zE
h i
þ e� lnð10Þm=z
�Ei
lnð10Þz
0:657m
� �
� Eilnð10Þ
zðmþ g� 80Þ
� ���¼ 1 ð10Þ
Each one of three terms in brackets in Eq. 10
corresponds to the heating cycle, the initial lag of the
cooling cycle, and the final, straight line, part of the cooling
cycle, respectively. The parameter g in Eq. 10 is the
difference between retort and product temperature, at the
critical point, at the end of heating,
Food Eng Rev (2010) 2:1–16 5
123
Page 6
g ¼ TRT � Tg ð11Þ
while the parameter m is the difference between product
temperature at the end of heating and retort temperature
during cooling, that is,
m ¼ Tg � TCW ð12Þ
The parameter U is the F value at retort temperature,
TRT, that is,
U ¼ FzTref
Fi ð13Þ
for
Fi ¼ 10Tref �TRT
z ð14Þ
The exponential integrals, E1(x) and Ei(x) are defined as
[16]:
E1ðxÞ ¼Z1
x
e�u
udu ð15Þ
and
EiðxÞ ¼Zx
�1
eu
udu ð16Þ
Values of the above exponential integrals can be
obtained through various approximations of Eqs. 15 and
16 or from tables [16]. The quantity E appearing in Eq. 10
is a numerical evaluated integral defined as:
E ¼Zlnð10Þ0:643m=z
lnð10Þ0:3m=z
e�u
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiu2 � ðlnð10Þ0:3m=zÞ2
qdu ð17Þ
There is an additional assumption hidden in Eq. 10. Due to
the limits of integration of Eq. 3, the initial product
temperature at the beginning of the heating cycle, as well as
the final product temperature at the end of the cooling cycle
should, somewhere, appear in Eq. 10. In order to have results
independent of these two temperatures, Ball assumed that
there is no lethal effect of the process for product temperatures
less than 80�F below the retort temperature during heating,
TRT. Thus, integration starts at TRT-80 (�F), giving the term
E1(ln(10)80/z) in Eq. 10, and ends up at when product
temperature reaches again TRT-80 (�F) during cooling, from
where the term Ei(ln(10)(m ? g–80)/z) in Eq. 10 comes
from. In consistence with all other temperature terms that
appear on Eq. 10 as differences between product and (heating
or cooling) retort temperature, note that the quantity m ? g–
80 is equal to TRT-80-TCW, that is the difference between
product temperature at the end of cooling (TRT-80) and retort
temperature during the cooling cycle (TCW).
At this point, we must mention that Eq. 10 is presented
here in its corrected form [46], which is slightly different
than the corresponding equations presented by Ball in his
first publication of the method [3] and by Ball and Olson in
their book on 1957 [6]. As a matter of fact, the equations
presented in these two publications, that is, in Ball [3] and in
Ball and Olson [6] are not only different than Eq. 10, but
they are also different between themselves [24, 46]. This
created some controversy [15, 43, 45] until finally being
resolved [29, 46]. Tabulated or graphically presented results
in all Ball’s publications dealing with the presentation of his
original formula method [3, 4, 6] are identical, but not in
accordance to the aforementioned two equations presented
in Ball [3] and Ball and Olson [6] publications [42]. Dis-
crepancies between Ball’s equation and the tabulated values
were attributed to typographical errors with the validity of
Ball’s tables being reestablished and the correct form of the
equation, as presented here by Eq. 10 being presented [46].
With the aforementioned remarks, and noting that Eq. 10
does not possess an analytical solution, after the graphical
evaluation of the integral E, Ball presented discrete values of
the solution of Eq. 10 in tabular or graphical forms [3, 4, 6].
His graphs or tables correlated fh/U values vs g or log(g)
having the z and the m ? g values as parameters. Ball pre-
sented results for three m ? g values, namely 130, 160 and
180�F, and 11 z values namely 6, 8, 10, 12, 14, 16, 18, 20, 22,
24 and 26�F. Such a graph, reproduced from the tabulated
values from [6], is presented here in Fig. 3. Note that Ball
used temperatures in degrees Fahrenheit which we kept in
this article too. Based on Eqs. 11 and 12, m ? g is defined as:
mþ g ¼ TRT � TCW ð18Þ
Minimum g values employed were equal to about 0.1�F
(log(g) = -1) as can be seen from Fig. 3 as well as Ball’s
graphs and tables [3, 4, 6]. For thermal processes with
g \ 0.1�F, Ball suggested to consider that the temperature
at the critical point remained constant and equal to TR -
0.1 (�F) from the moment the critical point has reached a
temperature of 0.1�F below retort temperature during
heating. Mathematically, for g \ 0.1�F, the correlation
between fh/U and log(g) is given by Eq. 19.
fhU
����g\0:1
¼100:1=z � fh
U
��g¼0:1
100:1=z � 1þ logðgÞð Þ � fhU
��g¼0:1
ð19Þ
where fhU
��g\0:1
is the fh/U value for the particular m ? g and
z value and for the g \ 0.1�F value of interest, and fhU
��g¼0:1
the corresponding fh/U value for g = 0.1�F, a value that
can be obtained directly from Ball’s graphs or tables for the
appropriate z value.
The above equation produces results absolutely equiv-
alent to the graphical data presented by NCA [32] in figures
where the log(g) range has being extended up to –8.
If another small enough g value, lower than or close to
6 Food Eng Rev (2010) 2:1–16
123
Page 7
0.1�F, g0, instead of the g = 0.1�F value is used from Ball’s
tables, then Eq. 19 reduces to:
fhU
����g\g0¼
10g0=z � fhU
��g¼g0
10g0=z � 1þ logðgÞð Þ � fhU
��g¼g0
ð20Þ
As mentioned earlier, the goal in thermal process
calculations is to obtain a relationship between heating
time and F value. The F value is directly correlated with
the fh/U variable through Eq. 12. The parameter g, the
difference between retort and product temperature (at the
critical point) at steam-off time, used as the independent
variable, was related to the end of heating (steam-off) time,
B, through the following formula (originated from Eq. 4):
B ¼ fhðlogðjhIÞ � logðgÞÞ ð21Þ
for
I ¼ TRT � TIT ð22Þ
Thus, through the solution of Eq. 10, and Eqs. 13 and
21, a relationship between heating time, B, and F value is
established and thermal process calculations can be
performed. In brief, the use of Ball’s method for
determining the F value of a given process, involves the
following steps:
1. Determination of the fh and jh values from the
experimental heating curve
2. Calculation of log(g) from Eq. 21 and the experimen-
tal steam-off time, B
3. Finding the fh/U value, for the log(g) value calculated
in step 2, from Ball’s graphs (for example, Fig. 3) and
the appropriate m ? g and z values
4. Calculation of F value (from the fh/U value found in
step 3) through the following equation (originated
from Eq. 13 -the definition of U):
FzTref¼ fh
fhUFi
ð23Þ
The second type of problem, that is, estimation of the
required heating (steam-off) time B in order to achieve a
particular Frequired value, requires the same steps, with the
last three steps being followed in reverse order.
Specifically, it involves the following:
1. Determination of the fh and jh values from the
experimental heating curve
2. Calculation of the fh/U value from the fh value found in
step 1 and the given Frequired value, through Eq. 13
3. Finding the log(g) value, for the fh/U value calculated
in step 2, from Ball’s graphs (for example, Fig. 3) and
the appropriate m ? g and z values
4. Calculation of the required steam-off time, B from
Eq. 21 and the log(g) found in step 3
Changes in retort or initial product temperatures do not
impose any particular problem with straight-line heating
curves since the fh and jh values are not affected.
Discussion
During the presentation of Ball’s formula method, we make
notice of a number of assumptions and simplifications that
were made in the development of the method. In the next
paragraphs we will try to address them.
Broken-Line Heating
The assumption about straight-line heating, that is, con-
stant fh throughout the entire heating cycle was in fact
handled by Ball. By providing additional tables (and
working equations), he allowed calculations for broken-
line heating curves, that is, heating curves that could not be
adequately described by a single straight line but needed a
0.1
1
10
100
1000
-1 -0.5 0 0.5 1 1.5 2
log(g) (g in °F)
fh/U
z = 6°F
z = 10°F
z = 14°F
z = 18°F
z = 22°F
z = 26°F
Fig. 3 Curves giving the fh/U vs log(g) relationship, according to
Eq. 10, for m ? g = 180�F and various z values (based on Ball and
Olson’s [6] data)
Food Eng Rev (2010) 2:1–16 7
123
Page 8
set of straight lines, each one with different slope, to be
described. For the case of a broken-line heating curve with
two straight lines of different slopes, fh1and fh2
(that is,
with one break), the following equation correlates the F
value with the fh/U parameter associated with Ball’s fh/U vs
log(g) graphs or tables:
fhU¼ fh2
FzTref
Fi þqbhðfh2
�fh1Þ
fhUjbh
ð24Þ
The fhU
��bh
parameter represents the fh/U value obtained
from Ball’s fh/U vs log(g) graphs or tables for g = gbh. The
parameter gbh is the temperature difference between
product (at the critical point) and retort temperature at
the time where the break, that is, the change in the slope of
the heating curve occurs. Similarly, qbh, is the q value for
g = gbh. During the calculations involved in solving
Eq. 10, the contribution of each phase of the process
(heating cycle, the initial lag of the cooling cycle, and the
final part of the cooling cycle) could be easily separated. So
Ball, based on these calculations, was able to estimate the
fraction, q, of the Fprocess value, which is achieved during
the heating cycle only of the thermal process, (assuming
fc = fh = constant). As earlier, for the fh/U vs g
relationship, Ball [3] initially presented graphs with
curves correlated q with g and later tables q vs g [6] for
a wider range of m ? g (130, 150, 160 and 180�F) and z
(6, 14, 18, and 26�F) values.
Thus, thermal process calculations for broken-line heat-
ing curves could be performed in an analogous way to
straight-line heating curves. The Fprocess value could be
directly calculated from Eq. 24, while the formula connected
heating time, B, with temperature difference at steam-off, g,
for broken-line heating curves, takes the form of:
B ¼ fh1logðjIÞ þ ðfh2
� fh1Þ logðgbhÞ � fh2
logðgÞ ð25Þ
Ball presented equations (similar to Eq. 24) considering
up to five different slopes (three for the heating curve and
two for the cooling curve) and in principle, one can extent
the methodology to handle any shape of heating and
cooling curves. This though is troublesome, subject to
errors, and probably the use of a General Method scheme is
more advisable.
In a last comment about broken-line heating curves, we
must mention that, contrary to straight-line heating curves
where changes in retort or initial product temperatures can
be easily handled, this is not the case for broken-line
heating. Since there is no method that provides a way of
estimating the time or the temperature where the brake
occurs for different processing conditions, it is considered
unrealistic to extrapolate heat penetration data to different
process conditions. Berry and Bush [8] suggested safe
procedures (which might give though rise to quality
problems) for extrapolating data to process with different
retort or initial product temperatures.
The fc = fh Case
A direct application of the broken-line heating curves
analysis is for the case where the slopes of the heating and
cooling curve are different. For such cases, the correlation
between the F value and the fh/U parameter is given by:
fhU¼ qfh þ ð1� qÞfc
FzTref
Fið26Þ
The existence in Ball’s formula method of the q vs g
correlations allows easy handling of the fc = fh case,
through the above equation.
Non Constant Retort Temperature
The equations presented so far assumed constant retort
temperature during heating, TRT, and during cooling, TCW.
Except of some cases of continuous processing, there is
some time period needed for the retort to achieve the con-
stant operating temperature during the heating phase, the so
called coming-up time (CUT). Based on experimental
observations, Ball concluded that a process at constant
retort temperature of duration equal to 42% of the CUT was
equivalent to the process at the variable retort temperature
profile during the CUT. So he suggested to correct the
heating time, by shifting the zero heating time axis by
0.58�CUT and apply his method, as for a process with
CUT = 0, based on this corrected ‘‘zero’’ axis. This cor-
rection factor has been widely used [31]. Obviously, the
percentage of the CUT that is lethal depends upon a number
of parameters, including the rate of change of retort tem-
perature until it reaches its operating value, as well as a
number of product characteristics and kinetic parameters of
the target microorganism [2, 7, 22, 39, 53, 56–58].
No similar correction was proposed for the time needed
for the retort to reach the cooling medium temperature.
Nevertheless, the assumption that retort reaches cooling
medium temperature, TCW, instantaneously after steam-off,
is a conservative approach.
Furthermore, no provisions were made in Ball’s formula
method for variable retort temperature processing. This
makes the method inadequate to compensate for possible
deviations, in terms of retort temperature, during
processing.
The jc Value of 1.41
Using a constant value for the cooling, lag factor jc is
probably the strictest assumption that Ball made in the
8 Food Eng Rev (2010) 2:1–16
123
Page 9
development of his formula method. As Ball [3] pointed
out: ‘‘… j, generally has a value different from 1.41. It is as
a rule less than this value.’’ In a next paragraph, he
somewhat implied to treat straight-line cooling curves, for
cases where jc = 1.41, as broken-line cooling curves. A
number of investigators, including Ball and Olson’s [6]
presentation of a new method, proposed equations and
presented tables allowing for variable jc values [18–20, 23,
25, 30, 33, 36–38, 51, 52].
With all other parameters being the same, using a jcvalue equal to 1.41 will be a conservative approach if the
true jc value is greater than 1.41. The opposite holds when
the true jc value is less than 1.41. A jc value of unity means
no lag in product temperature at the critical point, indi-
cating a uniform product temperature, as for the case of
complete mixed forced convection heated products. In
general, jc values less than 1 are unrealistic for the critical
point of the product.
A comparison of the effect of cooling jc values on the
fh/U vs log(g) relationship is presented in Fig. 4 with data
obtained from Stumbo [51] for z = 18�F. On the same
figure, Ball’s data for two m ? g values (130 and 180�F)
are included. Note that Stumbo [51] used a single,
constant m ? g value of 180�F to create his tables. Ball’s
values, for either m ? g value, are close to Stumbo’s data
for jc = 1.40, with Ball’s values producing slightly con-
servative results. Differences in Stumbo’s data for jcvalues of 1.00 and 1.40 are rather small, such as Stumbo’s
data for jc = 1.00 are reasonably approached by Ball’s
values. This can give an explanation of the successful
application of Ball’s method to forced convection heating
products. On the same line of thinking, the relative small
deviations between Ball’s and General method process
time calculations for thin profile packages reported by
Ghazala et al. [17] can be attributed to the jc values
involved in these cases, which were about 1.3, a value
close to 1.41, the one used by Ball. Differences in the
fh/U vs log(g) data observed between jc values of 2.00 and
1.40 produce conservative, in terms of safety, results
when Ball’s method is used. This though leads to over-
processing which can be detrimental to product quality.
Similar results were also obtained for all the other
z values used by Ball.
Temperature Rise After Steam-Off
As indicated earlier, the equation used to describe the
initial part of the cooling curve (Eq. 9) does not allow for
any product temperature rise at the beginning of cooling,
which is not valid for the critical point of conduction
heated canned foods (e.g., [13]. In such cases considerable
overprocessing can occur. As indicated in an earlier
publication [47], results obtained by Ball’s formula method
were closer to reference General Method values (in terms
of Fprocess calculations) for conduction heating products,
when the maximum product temperature, rather than the
product temperature at the end of heating, was considered
as Tg. As indicated then, Ball [3] actually defined g as ‘‘the
difference in degrees between the retort temperature and
the maximum temperature attained by the center of a can
during its processing’’. Similar definitions Ball gave in
subsequent publications [4, 6]. From the other hand, by the
definition of heating (process) time through Eq. 21, Ball
considered end of heating when the maximum product
temperature has been reached. This can be confusing, with
the recommendation to Ball’s formula method users being
to consider Tg as the critical point temperature at the end of
heating, allowing for any temperature rise after that to be
added as a safety factor [47]. The use of lower Frequired
values for larger cans of conduction heating foods is
believed to be an attempt to overcome the inability of
Ball’s method to allow for any temperature rise at the
critical point after steam-off [47].
0.1
1
10
100
1000
-1 -0.5 0 0.5 1 1.5 2
log(g) (g in °F)
fh/U
Ball m+g = 180°F
Ball m+g = 130°F
Stumbo jc = 0.40
Stumbo jc = 1.00
Stumbo jc = 1.40
Stumbo jc = 2.00
Fig. 4 Effect of cooling jc values on the fh/U vs log(g) relationship
for z = 18�F (based on Ball and Olson’s and Stumbo’s [6, 51] data)
Food Eng Rev (2010) 2:1–16 9
123
Page 10
Product Temperature at the Early Stage of Heating
Contrary to the approach that Ball followed to describe the
cooling curve, that is, the use of a separate equation to
describe the early phase of cooling (where product tem-
peratures are high), Ball used a single expression (Eq. 4)
for the entire heating curve. Equation 4 largely underesti-
mates product temperature at the early stage of heating.
However, for most conventional thermal processes, initial
product temperature is low enough so that there is no
destruction taking place at the beginning of the heating
and, therefore, the error in using Eq. 4 instead of a more
accurate one [20] is negligible. This might be though of
concern for aseptic processes [11] where high initial
product temperatures are encountered.
Product Temperatures 80�F Below TRT
As limits of integration in Eq. 3, Ball used a critical point
temperature 80�F below the retort temperature in both the
beginning of heating and at the end of cooling. He assumed
that there is no lethal effect of the process for product
temperatures less than 80�F below the retort temperature.
This is in general a valid assumption and can be prob-
lematic only if high z values are used. For z values up to
26�F that Ball used, this does not pose any problems.
Parametric Values
Ending our discussion concerning the validity of the
assumptions Ball made in developing his method, we must
emphasize the need of accurate knowledge of the various
parameters involved in the above calculations (for example
fh, jh, and z values). Deviation from first order kinetics and
the z value concept, as far as the effect of temperature on
the decimal reduction time is concerned, will require
reevaluation of Eq. 10 and therefore the whole method.
First order kinetics can be seen as a subset of different
inactivation schemes [59] and presumably, any thermal
process calculation method that is based upon such a model
will have increased flexibility and applicability. Statistical
variation in physical, reaction kinetics and operational
parameters can lead to large deviations in Fprocess, Frequired,
and heating time calculations. A comprehensive discussion
on this matter is given by [24].
Use of Algebraic Equations to Replace Ball’s Graphs
The use of Ball’s method relies on the tables or the graphs
with the discrete values of the solution of Eq. 10 that Ball
presented in the form of fh/U vs log(g) relationships. The
use of graphs or tables is in general difficult and susceptible
to errors. Several investigators tried to simplify the use of
the method through nomograms [35], computerized pro-
cedures [40] algebraic equations [60] or artificial neural
networks [1]. Use of algebraic equations in place of tabu-
lated fh/U vs log(g) values can ease the use of any thermal
process calculation method [26]. The polynomial equations
used by Vinters et al. [60] were though restricted to a z
value of 18�F. Thus, a new set of working, regression
equations were developed in order to facilitate the appli-
cation and use of Ball’s method, and they will be presented
in the rest of this paper.
The use of dimensionless variables can greatly reduce the
amount of tables, graphs or equations needed to express a
given relationship. Hayakawa [20] used a ratio of z values as
a working variable, Steele and Board [44] used ‘‘sterilization
ratios’’, that is, ratios of temperature differences between
product and heating (or cooling) medium over z values, while
[36, 37] used g/z ratios in his correlations for thermal process
calculations. A careful look at Eq. 10 reveals that tempera-
ture ratios of g/z and m/z are involved. If log(g/z) instead of
log(g) is used in the abscissa of the fh/U vs log(g) plots
(Fig. 3), one can see that curves for different z values almost
coincide. If a further transformation of the form of log(g/z)-
z/zc is used, curves for different z values become even closer,
as it is shown in Fig. 5 for the m ? g = 180�F data (same
data used for Fig. 3). For Fig. 5, a value of zc equal to 430�F
was used for demonstration purposes. A more accurate value
can be determined through a regression analysis and an
appropriate model.
Based on the above observation, an algebraic equation
correlating log(fh/U) (note the logarithmic scale used in
Fig. 3 and 5) with log(g/z)-z/zc was sought. After several
trials, the following equation is proposed to replace Ball’s
tabulated fh/U vs g values:
y ¼ a1
1þ a2e�a3xþ a4
1þ a5e�a6xþ a7 ð27Þ
The choice of Eq. 27 was made based on the residual
sum of squares error (SSE) and the correlation coefficient
(R2) between predicted and tabulated values (Table 1), and
the behavior of the proposed equation at the limits of the
variables and the parameters involved. Variables x and y
correspond to either log(fh/U) or log(g/z)-z/zc and vice
versa (Table 1) depending if we are seeking an fh/U value
for a given g, or the opposite. The nonlinear nature of the
correlation (Eq. 27) requires explicit algebraic equations
for both fh/U and g. The coefficients of Eq. 27 were
determined through a nonlinear regression using the
tabulated fh/U vs g data given by Ball and Olson [6] for
two m ? g values of 130�F and 180�F (Table 1). Although
data for m ? g = 160�F were also given in the original
publication [3], the fact that the m ? g parameter only
slightly affects the fh/U vs g relationship left us the option
10 Food Eng Rev (2010) 2:1–16
123
Page 11
of using only the two m ? g edge values. For every 10�F
change in the m ? g values, Stumbo [51] reported only an
error of 1% in the F value. An indication of the error
magnitude involved due to the m ? g variation can be
inferred from Fig. 4 by comparing fh/U vs g values for
m ? g = 130�F and m ? g = 180�F for z = 18�F. Note
that the use of the m ? g = 180�F data for any m ? g
value of less than 180�F represents a conservative
approach. Interpolation between the values obtained
through Eq. 27 for m ? g = 130�F and m ? g = 180�F
is recommended for intermediate m ? g values.
Note that the last term a7 in Eq. 27 was not necessary to
express the log(g/z) vs log(fh/U) relationship. All coeffi-
cients a1 through a7 appearing in Eq. 27 are dimensionless
with the exception of the zc coefficient which has units of
temperature difference, as the regular z value does. Thus,
temperature units in degrees Celsius can be used in Eq. 27
as long as the zc values presented on Table 1 will be
converted to temperature difference in degrees Celsius by
dividing the values given in Table 1 by 1.8.
Comparisons between predicted, through Eq. 27 and the
coefficients presented in Table 1, and Ball’s tabulated fh/U
vs log(g) data for m ? g = 180�F are presented in Fig. 6.
The agreement between the two data sets was very good.
The percent relative error between predicted and Ball’s
tabulated fh/U values was, with few exceptions,
within ±4%, and in the majority of the cases within ± 2%
(Fig. 7). This relative error in the fh/U values directly
reflects the relative error between Fprocess values calculated
through Eq. 27 or Ball’s tabulated fh/U vs log(g) data.
Similar results were obtained for the rest of the z values
for which Ball’s tabulated data were available, as well as
for the m ? g = 130�F data.
For the reverse calculations, that is, log(g) vs fh/U,
comparisons between predicted and Ball’s values revealed
also very good agreement, as shown for the m ? g =
130�F data which are indicatively presented in Fig. 8.
Absolute errors between predicted and Ball’s tabulated
log(g) values were less than ± 0.02 with the majority
being less than ± 0.01 (Fig. 9). In view of Eq. 21, the
absolute error in log(g) values is transferred through the
0.1
1
10
100
1000
-2.5 -2 -1.5 -1 -0.5 0 0.5
log(g/z)-z/zc
f h/U
z = 6°F
z = 10°F
z = 14°F
z = 18°F
z = 22°F
z = 26°F
Fig. 5 Re-plot of Fig. 3 data using a dimensionless log(g/z)-z/zc
variable (a zc value equal to 430�F is used)
Table 1 Values of the
coefficients of Eq. 27 according
to the definitions of the x and yvariables
m ? g = 130�F m ? g = 180�F
y log(g/z)-z/zc log(fh/U) log(g/z)-z/zc log(fh/U)
x log(fh/U) log(g/z)-z/zc log(fh/U) log(g/z)-z/zc
a1 -0.088335831 40.122199 -3.3545727 22.016510
a2 -0.96375429 38.533071 -0.34453049 21.598294
a3 0.028257272 2.3715954 0.42100067 2.4586869
a4 1.0711536 5.3058320 4.0057210 38.202986
a5 0.19518983 2.8885491 0.13211471 23.706331
a6 4.5699218 0.63534158 3.2971998 0.49435142
a7 - -0.63814873 – -0.74859566
zc (�F) 389.10600 405.49832 389.48491 468.11021
R2 0.999932 0.999951 0.999924 0.999954
SSE 0.0202 0.0202 0.0215 0.0185
Data points 584 584 578 578
Food Eng Rev (2010) 2:1–16 11
123
Page 12
fh value as an absolute error to the required heating time
calculations. Thus, a 0.01 absolute error in log(g) generates
a -0.01 9 fh absolute error in the heating time, B.
The use of Eq. 27 can largely facilitate Ball’s formula
method calculations. It cannot only make calculations easier
and with minimal error (compared to graphical data) but it
allows calculations for intermediate values of the parameters
involved; that is, for any intermediate fh/U and g (compared
to the tabulated data) as well as z values. It is worth men-
tioning that Eq. 27 produced reasonable predictions beyond
the z value of 26�F, which was the limit of Ball’s data, due to
its dimensionless form and the reasoning behind its selection
from a number of other regression equations. So, for
example, fh/U vs g values for z = 54�F through Eq. 27 were
almost identical to Stumbo’s [51] values for jc = 1.40.
Without recommending extrapolation of data, the above
observation is to support the value of Eq. 27.
Equation 27 is based on the tabulated fh/U vs g data
given by [6] where g values greater of about 0.1�F were
used. So, Eq. 27 is only valid for g C 0.1�F. For cases
where g is less than 0.1�F, Eq. 19 or 20 must be used
(either for fh/U or log(g)) with the fhU
��g¼0:1
or fhU
��g¼g0
values
needed for such calculations to be taken from Ball’s tables,
or determined through Eq. 27.
Ending our presentation in the use of algebraic equations
to replace Ball’s graphs and tables, and for the complete-
ness of calculations, the following equation, Eq. 28, was
developed to replace q vs g tabulated or graphical data
needed for broken-heating curves and/or the cases where
fc = fh.
q ¼ b1
1þ b2e�b3 logðgÞ þ1
1þ b4e�b5 logðgÞ þ ðb6gþ b7Þz
ð28Þ
Based on Ball and Olson [6] q vs g tables, the
coefficients of Eq. 28 were estimated and are presented
0.1
1
10
100
1000
-1 -0.5 0 0.5 1 1.5 2
log(g) (g in °F)
fh/U
z = 6°F
z = 10°F
z = 14°F
z = 18°F
z = 22°F
z = 26°F
Fig. 6 Comparison between predicted, through Eq. 27 and the
coefficients presented in Table 1, (lines) and Ball’s tabulated fh/Uvs log(g) data (open cycles) for m ? g = 180�F
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
00010010111.0
f h /U Ball
(fh/
UB
all-
f h/U
pred
icte
d)/
(fh/
Uba
ll)
(%
)
z = 6°F z = 10°F z = 14°F
z = 18°F z = 22°F z = 26°F
Fig. 7 Relative error between
predicted, through Eq. 27 and the
coefficients presented in Table 1, and
Ball’s tabulated fh/U data for
m ? g = 180�F
12 Food Eng Rev (2010) 2:1–16
123
Page 13
in Table 2, for m ? g = 130, 160 and 180�F. A
comparison between predicted, through Eq. 28 and the
coefficients presented in Table 2, and Ball’s tabulated q vs
g data for m ? g equal to 130 and 180�F are indicatively
presented in Fig. 10. As it can be seen (Fig. 10), the
agreement between the predicted and Ball’s data was very
-1
-0.5
0
0.5
1
1.5
2
0.10 1.00 10.00 100.00 1000.00
f h /U
log(
g) (
g in
°F
)
z = 26°F
z = 22°F
z = 18°F
z = 14°F
z = 10°F
z = 6°F
Fig. 8 Comparison between predicted, through Eq. 27 and the coefficients presented in Table 1, (lines) and Ball’s tabulated log(g) vs fh/U data
(open cycles) for m ? g = 130�F
-0.03
-0.02
-0.01
0.00
0.01
0.02
0.03
-1.0 -0.5 0.0 0.5 1.0 1.5
log(g) Ball (g in °F)
log(
g)B
all-
log(
g)pr
edic
ted
z = 6°F z = 10°F z = 14°F
z = 18°F z = 22°F z = 26°F
Fig. 9 Absolute error between predicted, through Eq. 27 and the coefficients presented in Table 1, and Ball’s tabulated log(g) data for
m ? g = 130�F
Food Eng Rev (2010) 2:1–16 13
123
Page 14
good. The maximum relative error between predicted and
Ball’s tabulated q values reported in Table 2, is rather
unjust for the correlation presented by Eq. 28, usually due
to some accidental extreme values.
Conclusions
Ball’s original formula method for thermal process calcu-
lations continues to serve the food industry since its
development in 1923. It represents an excellent example of
the intelligence use of mathematics in food processing.
In the preceding paragraphs, the fundamental ideas
behind thermal process calculations were initially out-
lined. Thereafter, the steps involved in the development
of Ball’s method were given with the appropriate clari-
fications. Key assumptions associated with the method
and their implications on product safety and quality were
discussed, and the conservative nature of the method was
pointed out.
Finally, a set of algebraic equations to replace Ball’s
fh/U vs log(g) tabulated or graphical data for g values greater
than 0.1�F, Eq. 27, and q vs log(g) data, Eq. 28, were pre-
sented. The use of these equations introduced negligible
error, can greatly facilitate the use of Ball’s original formula
method, and permit calculations for intermediate values of
the parameters involved. An explicit expression (Eq. 19 or
20) according to Ball’s assumptions, for the fh/U vs g rela-
tionship for g less than 0.1�F, was also given.
References
1. Afaghi MK (2000) Application of artificial neural network
modeling in thermal process calculations of canned foods. MSc
Thesis, Dept Food Science and Agr Chem, McGill University,
Montreal, Canada
2. Alstrand DV, Benjamin HA (1949) Thermal processing of canned
foods in tin containers. V. Effect of retorting procedures on
sterilization values in canned foods. Food Res 14:253–257
3. Ball CO (1923) Thermal process time for canned food. Bulletin
of the National Research Council No. 37., 7, Part 1Natl Res
Council, Washington, DC
4. Ball CO (1928) Mathematical solution of problems on thermal
processing of canned food. Univ Calif Pubs Public Health, vol 1,
No 2. University of California Press, Berkeley, CA (with sup-
plement in 1936)
5. Bal CO (1949) Process evaluation. Food Technol 3:116–118
6. Ball CO, Olson FCW (1957) Sterilization in food technology.
Theory practice and calculations. McGraw-Hill Book Co., New
York
7. Berry MR Jr (1983) Prediction of come-up time correction factors
for batch-type agitating and still retorts and the influence on
thermal process calculations. J Food Sci 48:1293–1299
8. Berry MR Jr, Bush RC (1987) Establishing thermal processes for
products with broken heating curves from data taken at other
retort and initial temperatures. J Food Sci 52(4):958–961
9. Bigelow WD, Bohart GS, Richardson AC, Ball CO (1920) Heat
penetration in processing canned foods. Bull. No. 16-L, Res. Lab.
Natl Canners Assoc, Washington, DC
10. Cleland AC, Robertson GL (1985) Determination of thermal
processes to ensure commercial sterility of foods in cans. In:
Thorne S (ed) Developments in food preservation, vol 3. Elsevier
Applied Science Publishers, London
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.1 1 10 100
log(g) (g in °F)
ρ
6°F 14°F 18°F 26°F
6°F 14°F 18°F 26°F
m + g = 180°F
m + g = 130°F
Fig. 10 Comparison between predicted, through Eq. 28 and the
coefficients presented in Table 2, (lines) and Ball’s tabulated q vs
log(g) data (open symbols) for m ? g = 130�F and for
m ? g = 180�F
Table 2 Values of the coefficients of Eq. 28
m ? g = 130�F m ? g = 160�F m ? g = 180�F
b1 0.289514 0.0552200 0.00993234
b2 -4.87991 -1.84861 -1.16901
b3 0.553369 0.240032 0.0647075
b4 0.0287250 0.0314127 0.0329320
b5 -1.89101 -1.79959 -1.72143
b6 -0.0000302250 -0.0000244004 -0.0000198010
b7 -0.000814476 -0.000521808 -0.000448516
R2 0.999928 0.999940 0.999958
SSE 0.000599 0.000362 0.000211
Data points 220 220 220
Maximum
relative error
2.65% 1.43% 0.52%
14 Food Eng Rev (2010) 2:1–16
123
Page 15
11. Dail R (1985) Calculation of required time of aseptically pro-
cessed low acid foods containing particulates utilizing the Ball
method. J Food Sci 50:1703–1706
12. Esty JR, Meyer KF (1922) The heat resistance of the spores of B.botulinus and allied anaerobes. J Infectious Dis 34(1):650–663
13. Flambert CMF, Deltour J (1971) Calcul des temperatures dans les
aliments en boıtes cylindriques transmettant la chaleur par con-
duction. Indust Alim Agr 88(9):1279–1287
14. Flambert F, Deltour J (1972) Localization of the critical area in
thermally-processed conduction heated canned foods. Lebensm
Wiss Technol 5(1):7–13
15. Flambert CMF, Deltour J, Dickerson RW, Hayakawa K (1977)
Lethal effect of food temperature on linear portion of a heating or
cooling curve. J Food Sci 42:545–546
16. Gautschi W, Cahill WF (1964) Exponential integral and related
functions. In: Abramowitz M, Stegun IA (eds) Handbook of
mathematical functions. Natl Bureau of Standards, Appl Math Ser
55, U.S. Covt Print Office, Washington, DC
17. Ghazala S, Ramaswamy HS, Smith J, Simpson B (1991) Thermal
process time calculations for thin profile packages: comparison of
formula methods. J Food Process Eng 13:269–282
18. Griffin RC, Jr, Herndon DH, Ball CO (1969) Use of computer
derived tables to calculate sterilizing processes for packaged
foods. 2. Application to broken line heating curves. Food Technol
23(4):519–524
19. Griffin RC Jr, Herndon DH, Ball CO (1971) Use of computer
derived tables to calculate sterilizing processes for packaged foods.
3. Application to cooling curves. Food Technol 25(2):134–143
20. Hayakawa K (1970) Experimental formulas for accurate esti-
mation of transient temperature of food and their application to
thermal process evaluation. Food Technol 24(12):1407–1418
21. Hayakawa K (1978) A critical review of mathematical procedures
for determining proper heat sterilization processes. Food Technol
32(3):59–65
22. Hayakawa K, Ball CO (1971) Theoretical formulas for temper-
atures in cans of solid food and for evaluating various heat pro-
cesses. J Food Sci 36:306–310
23. Herndon DH, Griffin RC Jr, Ball CO (1968) Use of computer
derived tables to calculate sterilizing processes for packaged
foods. Food Technol 22(4):473–484
24. Holdsworth D, Simpson R (2008) Thermal Processing of Pack-
aged Foods, 2nd edn. Springer, New York
25. Jen Y, Manson JE, Stumbo CR, Zahradnik JW (1971) A proce-
dure for estimating sterilization of and quality factor degradation
in thermally processed foods. J Food Sci 36:692–698
26. Kao J, Naveh D, Kopelman IJ, Pflug IJ (1981) Thermal process
calculations for different z and jc values using a hand held cal-
culator. J Food Sci 46:193–197
27. Kim KH, Teixeira AA, Bichier J, Tavares J (1993) STERILI-
MATE: software for designing and evaluating thermal steriliza-
tion processes. ASAE paper No 93–4051. St. Joseph, MI
28. Larkin JW (1989) Use of a modified Ball’s formula method to
evaluate aseptic processing of foods containing particulates. Food
Technol 43(3):124–131
29. Larkin JW (1990) Revision of the hyperbolic equation used by
the Ball formula method of thermal process evaluations. Paper
No 157, 50th Annual Meeting of Institute of Food Technologists,
June 16–20, Anaheim, CA
30. Larkin JW, Berry RB (1991) Estimating cooling process lethality
for different cooling j values. J Food Sci 56(4):1063–1067
31. Merson RL, Singh RP, Carroad PA (1978) An evaluation of
Ball’s formula method of thermal process calculations. Food
Technol 32(3):66–72
32. NCA (1968) Laboratory manual for food canners and processors,
Vol. 1. Microbiology and processing. Natl Canners Assoc Res
Labs The Avi Pub. Co., Inc., Westport, CT
33. Noronha J, Hendrickx M, Van Loey A, Tobback P (1995) New
semi-empirical approach to handle time-variable boundary con-
ditions during sterilisation of non-conductive heating foods.
J Food Eng 24:249–268
34. Olson FCW, Jackson JM (1942) Heating curves. Theory and
practical application. Ind Eng Chem 34(3):337–341
35. Olson FCW, Stevens HP (1939) Thermal processing of canned
foods in tin containers. II. Nomograms for graphic calculation of
thermal processes for non-acid canned foods exhibiting straight-
line semi-logarithmic heating curves. Food Res 4:1–20
36. Pham QT (1987) Calculation of thermal process lethality for
conduction heated canned foods. J Food Sci 52(4):967–974
37. Pham QT (1990) Lethality calculation for thermal processes with
different heating and cooling rates. Int J Food Sci Technol
25(2):148–156
38. Purohit KS, Stumbo CR (1973) Refinement and extension of
fh/U:g parameters for process calculation. J Food Sci 38:726–728
39. Ramaswamy HS (1993) Come-up time effectiveness for process
calculations involving thin-profile packages. J Food Eng 19:109–
117
40. Sasseen DM (1969) Computer program for process calculation by
the simplified Ball formula method. Natl Canners Assoc, West Br
Lab, Berkeley, CA
41. Simpson R, Almonacid S, Teixeira A (2003) Bigelow’s general
method revisited: development of a new calculation technique.
J Food Sci 68(4):1324–1333
42. Smith T, Tung MA (1982) Comparison of formula methods for
calculating thermal process lethality. J Food Sci 47:626–630
43. Steele RJ, Board PW (1979) Amendments to Ball’s formula
method for calculating the lethal value of thermal processes.
J Food Sci 44:292–293
44. Steele RJ, Board PW (1979) Thermal process calculations using
sterilizing ratios. J Food Technol 14:227–235
45. Steele RJ, Board PW, Bert DJ, Willcox ME (1979) Revision of
the formula method tables for thermal process evaluation. J Food
Sci 44:954–957
46. Stoforos NG (1991) On Ball’s formula method for thermal pro-
cess calculations. J Food Proc Eng 13(4):255–268
47. Stoforos NG (1995) Thermal process design. Food Control 6(2):
81–94
48. Stoforos NG (2003) Heating and cooling lag constants. In:
Heldman DR (ed) Encyclopedia of agricultural, food and bio-
logical engineering. Marcel Dekker, Inc., New York
49. Stoforos NG, Taoukis PS (2006) Heat processing: Temperature–
time combinations. In: Hui YH (ed) Handbook of food science,
technology, and engineering, vol 3. CRC Press, Taylor & Francis
Group, Boca Raton, FL
50. Stoforos NG, Noronha J, Hendrickx M, Tobback P (1997) A
critical analysis of mathematical procedures for the evaluation
and design of in-container thermal processes for foods. Crit Rev
Food Sci Nutr 37(5):411–441
51. Stumbo CR (1973) Thermobacteriology in food processing, 2nd
edn. Academic Press, Inc., New York
52. Stumbo CR, Longley RE (1966) New parameters for process
calculation. Food Technol 20(3):341–345
53. Succar J, Hayakawa K (1982) Prediction of time correction factor
for come-up heating of packaged liquid food. J Food Sci 47:614–
618
54. Teixeira AA, Dixon JR, Zahradnik JW, Zinsmeister GE (1969)
Computer determination of spore survival distributions in ther-
mally-processed conduction-heated foods. Food Technol 23(3):
78–80
55. Teixeira AA, Balaban MO, Germer SPM, Sadahira MS, Te-
ixeire Neto RO, Vitali AA (1999) Heat transfer model perfor-
mance in simulation of process deviations. J Food Sci 64(3):
488–493
Food Eng Rev (2010) 2:1–16 15
123
Page 16
56. Townsend CT, Reed JM, McConnell J, Powers MJ, Esselen WB
Jr, Somers II, Dwyer JJ, Ball CO (1949) Comparative heat pen-
etration studies on jars and cans. Food Technol 3(6):213–226
57. Uno J, Hayakawa K (1980) Correction factor of come-up heating
based on critical point in a cylindrical can of heat conduction
food. J Food Sci 45(4):853–859
58. Uno J, Hayakawa K (1981) Correction factor of come-up heating
based on mass average survivor concentration in a cylindrical can
of heat conduction food. J Food Sci 46:1484–1497
59. van Boekel MAJS (2002) On the use of the Weibull model to
describe thermal inactivation of microbial vegetative cells. Int J
Food Microbiol 74:139–159
60. Vinters JE, Patel RH, Halaby GA (1975) Thermal process eval-
uation by programmable computer calculator. Food Technol
29(3):42,44,46,48
16 Food Eng Rev (2010) 2:1–16
123